GitHub-Proxy/eigen
1
0
Fork 0
mirror of https://gitlab.com/libeigen/eigen.git synced 2025-09-22 22:33:15 +08:00
Code Issues Packages Projects Releases Wiki Activity

Now testing out (+=, =) in.FUNC() and out (+=, =) out.FUNC()

Browse Source
This commit is contained in:
Luke Iwanski 2016-11-18 11:16:42 +00:00
parent 7c30078b9f
commit 927bd62d2a
1 changed files with 52 additions and 22 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

74
unsupported/test/cxx11_tensor_builtins_sycl.cpp
View File

@ -32,11 +32,14 @@ template <typename T> T cube(T x) { return x * x * x; }
template <typename T> T inverse(T x) { return 1 / x; } template <typename T> T inverse(T x) { return 1 / x; }
} }
#define TEST_UNARY_BUILTINS_FOR_SCALAR(FUNC, SCALAR) \ #define TEST_UNARY_BUILTINS_FOR_SCALAR(FUNC, SCALAR, OPERATOR) \
{ \ { \
/* out OPERATOR in.FUNC() */ \
Tensor<SCALAR, 3> in(tensorRange); \ Tensor<SCALAR, 3> in(tensorRange); \
Tensor<SCALAR, 3> out(tensorRange); \ Tensor<SCALAR, 3> out(tensorRange); \
in = in.random() + static_cast<SCALAR>(0.01); \ in = in.random() + static_cast<SCALAR>(0.01); \
out = out.random() + static_cast<SCALAR>(0.01); \
Tensor<SCALAR, 3> reference(out); \
SCALAR *gpu_data = static_cast<SCALAR *>( \ SCALAR *gpu_data = static_cast<SCALAR *>( \
sycl_device.allocate(in.size() * sizeof(SCALAR))); \ sycl_device.allocate(in.size() * sizeof(SCALAR))); \
SCALAR *gpu_data_out = static_cast<SCALAR *>( \ SCALAR *gpu_data_out = static_cast<SCALAR *>( \
@ -45,41 +48,68 @@ template <typename T> T inverse(T x) { return 1 / x; }
TensorMap<Tensor<SCALAR, 3>> gpu_out(gpu_data_out, tensorRange); \ TensorMap<Tensor<SCALAR, 3>> gpu_out(gpu_data_out, tensorRange); \
sycl_device.memcpyHostToDevice(gpu_data, in.data(), \ sycl_device.memcpyHostToDevice(gpu_data, in.data(), \
(in.size()) * sizeof(SCALAR)); \ (in.size()) * sizeof(SCALAR)); \
gpu_out.device(sycl_device) = gpu.FUNC(); \ sycl_device.memcpyHostToDevice(gpu_data_out, out.data(), \
(out.size()) * sizeof(SCALAR)); \
gpu_out.device(sycl_device) OPERATOR gpu.FUNC(); \
sycl_device.memcpyDeviceToHost(out.data(), gpu_data_out, \ sycl_device.memcpyDeviceToHost(out.data(), gpu_data_out, \
(out.size()) * sizeof(SCALAR)); \ (out.size()) * sizeof(SCALAR)); \
for (int i = 0; i < in.size(); ++i) { \ for (int i = 0; i < out.size(); ++i) { \
VERIFY_IS_APPROX(out(i), std::FUNC(in(i))); \ SCALAR ver = reference(i); \
ver OPERATOR std::FUNC(in(i)); \
VERIFY_IS_APPROX(out(i), ver); \
} \ } \
sycl_device.deallocate(gpu_data); \ sycl_device.deallocate(gpu_data); \
sycl_device.deallocate(gpu_data_out); \ sycl_device.deallocate(gpu_data_out); \
} \
{ \
/* out OPERATOR out.FUNC() */ \
Tensor<SCALAR, 3> out(tensorRange); \
out = out.random() + static_cast<SCALAR>(0.01); \
Tensor<SCALAR, 3> reference(out); \
SCALAR *gpu_data_out = static_cast<SCALAR *>( \
sycl_device.allocate(out.size() * sizeof(SCALAR))); \
TensorMap<Tensor<SCALAR, 3>> gpu_out(gpu_data_out, tensorRange); \
sycl_device.memcpyHostToDevice(gpu_data_out, out.data(), \
(out.size()) * sizeof(SCALAR)); \
gpu_out.device(sycl_device) OPERATOR gpu_out.FUNC(); \
sycl_device.memcpyDeviceToHost(out.data(), gpu_data_out, \
(out.size()) * sizeof(SCALAR)); \
for (int i = 0; i < out.size(); ++i) { \
SCALAR ver = reference(i); \
ver OPERATOR std::FUNC(reference(i)); \
VERIFY_IS_APPROX(out(i), ver); \
} \
sycl_device.deallocate(gpu_data_out); \
} }
#define TEST_UNARY_BUILTINS_OPERATOR(SCALAR, OPERATOR) \
TEST_UNARY_BUILTINS_FOR_SCALAR(abs, SCALAR, OPERATOR) \
TEST_UNARY_BUILTINS_FOR_SCALAR(sqrt, SCALAR, OPERATOR) \
TEST_UNARY_BUILTINS_FOR_SCALAR(rsqrt, SCALAR, OPERATOR) \
TEST_UNARY_BUILTINS_FOR_SCALAR(square, SCALAR, OPERATOR) \
TEST_UNARY_BUILTINS_FOR_SCALAR(cube, SCALAR, OPERATOR) \
TEST_UNARY_BUILTINS_FOR_SCALAR(inverse, SCALAR, OPERATOR) \
TEST_UNARY_BUILTINS_FOR_SCALAR(tanh, SCALAR, OPERATOR) \
TEST_UNARY_BUILTINS_FOR_SCALAR(exp, SCALAR, OPERATOR) \
TEST_UNARY_BUILTINS_FOR_SCALAR(log, SCALAR, OPERATOR) \
TEST_UNARY_BUILTINS_FOR_SCALAR(abs, SCALAR, OPERATOR) \
TEST_UNARY_BUILTINS_FOR_SCALAR(ceil, SCALAR, OPERATOR) \
TEST_UNARY_BUILTINS_FOR_SCALAR(floor, SCALAR, OPERATOR) \
TEST_UNARY_BUILTINS_FOR_SCALAR(round, SCALAR, OPERATOR) \
TEST_UNARY_BUILTINS_FOR_SCALAR(log1p, SCALAR, OPERATOR)
#define TEST_UNARY_BUILTINS(SCALAR) \ #define TEST_UNARY_BUILTINS(SCALAR) \
TEST_UNARY_BUILTINS_FOR_SCALAR(abs, SCALAR) \ TEST_UNARY_BUILTINS_OPERATOR(SCALAR, += ) \
TEST_UNARY_BUILTINS_FOR_SCALAR(sqrt, SCALAR) \ TEST_UNARY_BUILTINS_OPERATOR(SCALAR, = )
TEST_UNARY_BUILTINS_FOR_SCALAR(rsqrt, SCALAR) \
TEST_UNARY_BUILTINS_FOR_SCALAR(square, SCALAR) \
TEST_UNARY_BUILTINS_FOR_SCALAR(cube, SCALAR) \
TEST_UNARY_BUILTINS_FOR_SCALAR(inverse, SCALAR) \
TEST_UNARY_BUILTINS_FOR_SCALAR(tanh, SCALAR) \
TEST_UNARY_BUILTINS_FOR_SCALAR(exp, SCALAR) \
TEST_UNARY_BUILTINS_FOR_SCALAR(log, SCALAR) \
TEST_UNARY_BUILTINS_FOR_SCALAR(abs, SCALAR) \
TEST_UNARY_BUILTINS_FOR_SCALAR(ceil, SCALAR) \
TEST_UNARY_BUILTINS_FOR_SCALAR(floor, SCALAR) \
TEST_UNARY_BUILTINS_FOR_SCALAR(round, SCALAR) \
TEST_UNARY_BUILTINS_FOR_SCALAR(log1p, SCALAR)
static void test_builtin_unary_sycl(const Eigen::SyclDevice &sycl_device) { static void test_builtin_unary_sycl(const Eigen::SyclDevice &sycl_device) {
int sizeDim1 = 100; int sizeDim1 = 10;
int sizeDim2 = 100; int sizeDim2 = 10;
int sizeDim3 = 100; int sizeDim3 = 10;
array<int, 3> tensorRange = {{sizeDim1, sizeDim2, sizeDim3}}; array<int, 3> tensorRange = {{sizeDim1, sizeDim2, sizeDim3}};
TEST_UNARY_BUILTINS(float) TEST_UNARY_BUILTINS(float)
TEST_UNARY_BUILTINS(double) TEST_UNARY_BUILTINS(double)
} }
void test_cxx11_tensor_builtins_sycl() { void test_cxx11_tensor_builtins_sycl() {